CANCELLED: Words to Swords: Translating Text into Actions - DLS Talk by Regina Barzilay, MIT

In this talk, I will address the task of using natural language instructions to guide control applications, such as game playing, code generation, and robot navigation. We assume access to natural language documents that describe the desired behavior of a control algorithm (such as game strategy guides). First, I will present a reinforcement learning algorithm for learning to map natural language instructions to executable actions. This technique has enabled automation of tasks that until now have required human participation — for example, automatically configuring software by consulting how-to guides and generating C++ code for input parsers. Next, I will present a Monte-Carlo search algorithm for game playing that incorporates information from game strategy guides. In this framework, the task of text interpretation is formulated as a probabilistic model that is trained based on feedback from Monte-Carlo search. When applied to the Civilization strategy game, a language-empowered player outperforms its traditional counterpart by a significant margin.